The TinMan AI Builder is a platform and
development kit for creating autonomous artificial intelligence and
real-time decision making systems. It includes an integrated
development environment application (IDE) and the runtime engine
(sold separately).
Most AI tools and AI middleware applications
today are impractical for development of an AI system that can
function autonomously in a dynamic environment. Many applications
require significant familiarity and/or programming skills to be
utilized properly, require a heavy mathematics or statistical
background to be productive or deal with a single problem domain.
TinMan provides a development environment and
underlying technology for the rapid production of complex
multi-module decision making systems ready for deployment in a
dynamic environment, abstracting and shielding the user from the
unnecessary mathematical tedium associated with these technologies.
The TinMan IDE allows the construction,
testing and simulation of an AI system, along with the packaging and
export of that system for use in the runtime environment. The
runtime environment provides a dynamic link library which loads,
reconstructs and then executes all AI systems and coordinates the
communication between and among the AI entities and their human
controllers, if any.
AI systems developed in TinMan do not have to
be autonomous, nor function in a continuous mode. An AI system can
be designed to receive an input data set and compute one or more
answers – from a single manual user selection event in the host
application. However, TinMan AI systems are extremely well-suited
for continuous processing of input data in a dynamic environment,
allowing for the continuous potentially perfect execution of all
desired outputs.
The process of constructing an AI system
involves the declaration of a set of variables, a set of outputs and
the selection and connection of a set of components. Over 85 types
of component templates are available for drag and drop addition to
the AI system, in familiar logic flow model. States of decision
making can either be made simultaneous or exclusive, as well as the
end actions from the system.
As variables and actions are defined and then
connected to the components, and as components are connected to each other,
appropriate structural modifications are automatically and
transparently made by the TinMan IDE. Although an AI system can be
viewed at the component level for a module or the entire system, the
composition and connectivity of the components is completely
abstracted.
With the logical design of a system complete
and connections and appropriate interpretations of data set,
thorough testing and simulation is performed. The resulting trained system is then
packaged and made ready for integration into the host application.
At runtime, the host application loads the
packaged AI system via the TinMan runtime engine, feeds the AI
system the set of input variable values, the system processes the
data via its configured module(s) and updates all system outputs
with updated values.
The
Build and Deploy Process
The TinMan approach to building and deploying
an AI system is designed to be efficient and straight-forward.
TinMan AI Builder provides the environment, technology and tools to
create and package for runtime, an artificial intelligence core (AI
core). The TinMan AI Runtime Library is integrated with a host
application to load and utilize that AI core to make runtime
decisions in the host system.
Runtime Execution Process
At runtime, the host application utilizes the
AI Core through the use of 3 simple steps (each with a single API
call). At each instance of time on a continuous basis, (or in
response to a single manual event in the host application):
1)
Update the values of
variables.
·
This is the information the AI System
needs to compute its response. This information may in some cases
come from physical sensors on-board the host system, or any other
source.
2)
Execute the AI System
·
This function feeds all variables to
the AI System, executes the system, and returns the multi-action
response pointer filled with all actions (by integer ID) resulting
from this cycle of logic.
3)
Act on the returned
action identifiers.
·
The host application simply reads and acts on each
system output updated following each system execution request.
Middleware Integration
Because of the simplicity and openness of
this approach, TinMan AI systems can easily and conveniently
integrate with other software systems, databases or specialized AI
middleware to best support the operation of the host system. For
instance a returned value from the execute API call above may be
asking for the calculation of a new 3-D path (from a path-finding
application), or to deliver a message wirelessly to a recipient
system, or to pull another record from a database for processing.
|
Integration of Middleware
Technologies
(hypothetical examples) |
|
